SAS Banking Predictions: AI Agents Handle Compliance and Fraud

SAS Maps the Future of Banking AI

SAS, the analytics and AI company that has served the banking industry for over four decades, has released its annual banking AI predictions report for 2026. The report compiles insights from 13 industry experts across banking, technology, and regulation to paint a picture of how AI agents will reshape financial services this year. The overarching theme is unmistakable: agentic AI is moving from experimental to operational across the most critical functions in banking, including compliance, fraud prevention, and customer engagement.

The timing of the report matters. Banks are under unprecedented pressure from multiple directions. Regulatory requirements have expanded significantly, with new anti-money laundering rules, consumer protection mandates, and data privacy obligations layering on top of existing Basel III and stress testing requirements. Simultaneously, fintech competitors continue to capture market share with superior customer experiences. AI agents offer banks the ability to meet regulatory obligations and competitive challenges simultaneously.

AI Agents for Compliance Monitoring

Compliance is the single largest operational cost center for most banks. JP Morgan alone spends an estimated $15 billion annually on regulatory compliance and risk management. SAS experts predict that 2026 will be the year AI agents transform compliance from a cost center into a competitive advantage.

flowchart LR
    REQ(["Inbound request"])
    PII["PII detection<br/>regex plus NER"]
    POL{"Policy engine<br/>OPA or rules"}
    REDACT["Redact or mask"]
    LLM["LLM call"]
    OUT["Response"]
    AUDIT[("Append only<br/>audit log")]
    BLOCK(["Block plus<br/>notify DPO"])
    REQ --> PII --> POL
    POL -->|Allow| REDACT --> LLM --> OUT --> AUDIT
    POL -->|Deny| BLOCK
    style POL fill:#4f46e5,stroke:#4338ca,color:#fff
    style AUDIT fill:#ede9fe,stroke:#7c3aed,color:#1e1b4b
    style BLOCK fill:#dc2626,stroke:#b91c1c,color:#fff
    style OUT fill:#059669,stroke:#047857,color:#fff

Continuous Regulatory Monitoring

Regulatory change is constant. Banks in the United States must comply with rules from the OCC, FDIC, Federal Reserve, CFPB, FinCEN, SEC, and state regulators, among others. Globally operating banks add dozens more regulatory bodies. AI agents now handle the continuous monitoring of regulatory publications, enforcement actions, and guidance updates across these agencies:

• Automated impact assessment: When a regulator issues new guidance, AI agents analyze the bank's current policies, procedures, and systems to identify gaps and quantify the effort required for compliance
• Policy drafting and updating: Agents generate draft policy updates that address new regulatory requirements, referencing the specific regulatory text and mapping changes to existing policy frameworks
• Control testing automation: Agents continuously test compliance controls by analyzing transaction data, document workflows, and employee actions to verify that controls are operating as designed

AML and KYC Agent Systems

Anti-money laundering and know-your-customer processes represent the compliance functions most immediately transformed by agentic AI. Traditional AML systems generate massive volumes of alerts, with false positive rates frequently exceeding 95 percent. Compliance analysts spend the majority of their time investigating alerts that turn out to be legitimate activity.

SAS experts predict that AI agents will reduce false positive rates to below 50 percent in 2026 at banks that deploy agentic systems, while simultaneously improving detection of genuine suspicious activity. This is achieved through:

• Contextual transaction analysis: Agents analyze each flagged transaction in the context of the customer's complete transaction history, peer group behavior, geographic patterns, and known typologies rather than applying simple threshold rules
• Entity resolution and network analysis: Agents map relationships between entities, accounts, and transactions to identify networks of suspicious activity that would be invisible when analyzing individual transactions in isolation
• Automated investigation workflows: When an agent determines that an alert requires investigation, it gathers relevant documentation, customer information, and transaction histories into a structured investigation package, reducing analyst time from hours to minutes per case
• Suspicious activity report drafting: For confirmed suspicious activity, agents draft the narrative sections of suspicious activity reports (SARs) based on the investigation findings, maintaining consistency and quality across filings

AI Agents for Fraud Prevention

Fraud losses in banking continue to escalate, with global losses exceeding $48 billion in 2025. SAS experts identify several areas where AI agents will advance fraud prevention in 2026:

Real-Time Fraud Triage

Traditional fraud detection systems produce a risk score for each transaction and route scores above a threshold to a fraud analyst queue. AI agents improve on this model by triaging alerts autonomously:

• Automated decisioning for clear cases: Agents approve or block transactions where the risk assessment is highly confident, reducing analyst workload by 40 to 60 percent while maintaining or improving detection rates
• Customer contact orchestration: For ambiguous transactions, agents initiate real-time customer verification through the customer's preferred channel, whether push notification, SMS, or phone call, resolving cases without analyst intervention
• Adaptive learning: Agents continuously learn from confirmed fraud cases and false positive investigations, updating their decision models to reflect evolving fraud patterns and customer behavior

Emerging Fraud Typology Detection

Fraud evolves constantly. New techniques such as deepfake-assisted social engineering, authorized push payment fraud, and synthetic identity fraud require detection approaches that can identify novel patterns rather than relying on known signatures. AI agents address this through anomaly detection that identifies transactions or behaviors that deviate from established patterns, even when the specific fraud technique has not been seen before.

Customer Onboarding Automation

The customer onboarding experience is a critical competitive battleground for banks. Traditional onboarding for a business banking account can take days or weeks, involving multiple document submissions, manual identity verification, and compliance checks. SAS experts predict that AI agents will compress this process to minutes for standard cases:

• Document processing agents: Agents extract and validate information from identity documents, corporate filings, financial statements, and other onboarding materials using advanced document understanding models
• Identity verification orchestration: Agents coordinate verification steps including document authentication, biometric matching, sanctions screening, adverse media checks, and credit bureau inquiries in parallel rather than sequentially
• Risk-based onboarding paths: Agents assess the risk profile of each applicant and dynamically adjust the onboarding requirements. Low-risk individuals receive streamlined processes while high-risk applicants are routed to enhanced due diligence
• Account configuration: Once onboarding checks are complete, agents configure the new account with appropriate products, limits, and services based on the customer's needs and risk profile

Regulatory Reporting Transformation

SAS experts highlight regulatory reporting as a function ripe for agent-driven transformation. Banks submit thousands of regulatory reports annually, each requiring data extraction from multiple source systems, calculation of specified metrics, validation against regulatory rules, and formatting according to specific templates. Errors in regulatory reports can result in fines, reputational damage, and increased supervisory scrutiny.

AI agents are being deployed to handle the end-to-end reporting pipeline: extracting data from source systems, performing calculations, running validation checks, generating reports, and flagging anomalies for human review before submission. This reduces report preparation time by 60 to 80 percent while improving accuracy.

The Banking AI Transformation Roadmap

SAS experts collectively outline a transformation roadmap that most banks are following or should follow:

• Phase 1: Augmentation: Deploy AI agents that assist human analysts in compliance, fraud, and customer service by providing recommendations, pre-processing data, and generating draft outputs. Most large banks are in this phase today
• Phase 2: Automation of routine decisions: Expand agent authority to make routine decisions autonomously, such as approving low-risk transactions, resolving clear false positive alerts, and completing standard onboarding checks. Leading banks are entering this phase in 2026
• Phase 3: Orchestration: Deploy multi-agent systems where specialized agents coordinate to handle complex, cross-functional processes end-to-end, such as the complete lifecycle of a suspicious activity investigation from detection to SAR filing. This phase is expected to mature in 2027 and 2028
• Phase 4: Autonomous operations: Full autonomous handling of entire operational domains, with human oversight focused on exception handling, strategy, and governance. This phase remains aspirational for most banks

Frequently Asked Questions

How do AI agents reduce AML false positive rates while improving detection?

Traditional AML systems use simple rules such as transaction amount thresholds or geographic flags that generate alerts regardless of context. AI agents analyze each transaction in the full context of the customer's behavioral history, peer group patterns, business type, and known fraud typologies. This contextual analysis allows agents to dismiss alerts that are clearly consistent with normal behavior while identifying genuinely suspicious patterns that rules-based systems miss. The net result is fewer false positives and better detection of real threats.

Are regulators comfortable with AI agents making compliance decisions?

Regulatory comfort varies by jurisdiction and function. Most regulators accept AI-assisted compliance as long as human oversight is maintained for significant decisions, the AI systems are explainable and auditable, and the bank can demonstrate that AI-assisted processes produce outcomes at least as good as human-only processes. Regulators in the US, UK, and EU have all published guidance that encourages responsible AI adoption in banking while emphasizing accountability and governance requirements.

What infrastructure do banks need to deploy AI agents for compliance and fraud?

Banks need a unified data layer that brings together transaction data, customer data, and external data sources in real time. They need model serving infrastructure capable of low-latency inference for real-time decisioning. They need workflow orchestration platforms that can route agent decisions and escalations appropriately. And they need comprehensive logging and audit trail capabilities to satisfy regulatory requirements. Most large banks have the foundational infrastructure but need to modernize data pipelines and add real-time processing capabilities.

What is the expected cost savings from deploying AI agents in banking compliance?

SAS experts estimate that banks deploying AI agents across AML, KYC, and regulatory reporting functions can reduce compliance operational costs by 25 to 40 percent within 18 to 24 months of production deployment. The savings come primarily from reduced analyst headcount needs for routine alert triage, faster investigation cycles, and automated report generation. However, upfront investment in technology, data infrastructure, and change management is significant, and ROI timelines vary based on the bank's starting point and scale.